Model systems for experimental studies: retinal ganglion cells in culture

The human retina is composed of nine different layers. The innermost layers contain the retinal ganglion cells (RGCs) and their axons in the ganglion cell layer and the nerve fiber layer, respectively. RGCs receive afferents from bipolar and amacrine cells and transmit efferents via action potentials to the brain, specifically the lateral geniculate nucleus, the superior colliculus, the pretectal nuclei, and the suprachiasmatic nucleus.

Optic neuropathies are diseases of the RGC and its axon. The most common optic neuropathy is glaucoma. Optic nerve diseases can be studied in vitro and in vivo using experimental models that range greatly in their applicable to glaucomatous optic neuropathy. There is an approximate hierarchy of models (Fig. 1), where the models higher

Corresponding author. Tel.: +1 514 252 3400; Fax: +1 514 251 7094

in hierarchy present an increased similarity to human optic neuropathies and the models lower in the pyramid allow better determination of the mechanisms responsible for the disease. This chapter discusses in vitro methods for studying RGCs (Levin, 2005). There are several types of culture models used for studying the pathophysiology of RGCs: (1) dissociated retinal cells, where the RGCs are either mixed with other cells or identified by labeling; (2) RGCs purified either by immunoaffinity techniques or differential centrifugation;

(3) retinal explants; (4) glial and other supporting cell cultures; and (5) RGC-like cell lines.

Mixed RGCs in culture

For mixed primary retinal cultures, neonatal or adult retinas can be dissociated enzymatically and maintained in culture for up to several weeks.

280

Fig. 1. In vitro and in vivo models of glaucoma can be viewed in a hierarchical pyramid, with the models at the top of the pyramid being more similar to the human disease, but less helpful in studying pathophysiology because of their complexity and experimental limitations. Models toward the bottom of the pyramid do not share clinical features of human glaucoma, but are more useful for studying mechanisms and screening potential therapies (Levin, 2005).

Thy-1. This methodology was first used by Barres using panning techniques (Barres et al., 1988), and it is the most commonly used method for studying isolated RGCs in vitro. Subsequently, other methods for purification have been developed, including the use of magnetic beads (Tezel and Wax, 2000).

Purified RGC cultures offer the advantage of studying the RGC in isolation without the effects of interactions with other cell types in the retina. RGC purity can be 95% or greater. However, the purification process requires exposure of RGC to antibodies to Thy-1, which might have biologic effects. Growth factors are typically used to maintain long term survival, for up to several weeks (Meyer-Franke et al., 1995).

RGCs can then be identified by different techniques. The most commonly used is retrograde labeling with a fluorescent dye. RGC can also be identified by immunolabeling against Thy-1 or Brn-3 (Leifer et al., 1984; Garcia et al., 2002; Leahy et al., 2004) or by using an RGC-specific promoter to drive a reporter gene if cultured from a transgenic animal (Feng et al., 2000). This type of culture allows the observation of interactions among a variety of cell types and the RGC because the diversity is maintained. However, the anatomic arrangement of the retina is not maintained. Axotomy is also inherent to this model because the optic nerve is cut when the eye is taken from the animal, and then the proximal RGC axon is cut when the retina is enzymatically dissociated. Survival is typically short as growth factors are typically not used besides those contained in serum.

Purified ganglion cells

In purified RGC cultures, enzymatic dissociation is performed and then the RGCs are isolated from other retinal neurons via immunoselection. Macrophages are typically removed first with antibodies followed by antibodies to the cell surface marker

Retinal explants

Retinal explant cultures can be made by dissection of whole retinas into small pieces (Smalheiser et al., 1981). The explant allows the maintenance of cellular diversity and anatomic arrangements. The injury effects of dissociation are avoided, but axotomy is still present, although not as close to the cell body as with cell dissociation.

A commonly used technique for explants was first used to demonstrate that RGC neurites from both embryonic and adult mice can grow on laminin, but adding antibodies to the b1/b3 integrin blocked the laminin-dependent growth on embryonic optic fibers only (Bates and Meyer, 1997). Subsequent studies from several laboratories have established the use of this technique for assessing neurite extension from RGCs (Bahr et al., 1988; Manabe et al., 2002), measuring RGC survival (Turner, 1985; Fischer et al., 2000; Xin et al., 2007), and studying RGC differentiation (Wang et al., 2002).

Glial cultures

Glial and other supporting cell types have also been used in culture. Retinal astrocytes and Mu¨ller cells can be cultured to varying degree of purity

as can optic nerve head lamina cribrosa cells (Hernandez et al., 1988).

These different cell populations can be studied with respect to neurotrophin secretion (Lambert et al., 2001), uptake of glutamate (Kawasaki et al., 2000), induction of injury signals (Neufeld et al., 1997), and other interactions with retinal neurons.

RGC-5 cells

The RGC-5 cell line is a transformed retinal ganglion cell. It was derived by transforming postnatal day 1 rat retinal cells by Krishnamoorthy et al. (2001). This cell line expresses neuronal markers’ characteristic of RGCs such as Thy-1, Brn-3, neuritin, synaptophysin, NMDAR1, and GABAB receptors. These cells are serumand neurotrophin-dependent. They do not express the astrocyte marker GFAP.

A great advantage of using these cells is that they are uniform in phenotype, allowing good repeatability of experiments. Also, as a dividing cell line, they are in principle of limitless availability. However, a disadvantage is that these cells are mitotically active and are therefore phenotypically different from a normal postmitotic RGC. In addition, RGC-5 cells are morphologically more similar to glial cells in culture than to primary RGC, and do not express the repertoire of ion channels characteristic of RGCs.

Differentiation of RGC-5 cells

RGC-5 cells can be treated with agents that differentiate them. Differentiation with succinyl concanavalin A (sConA) makes RGC-5 cells sensitive to glutamate toxicity. This glutamate excitotoxicity is blocked by NMDA antagonists (Krishnamoorthy et al., 2001). However, RGC-5 cells differentiated with sConA do not assume a neuronal morphology, nor is proliferation halted.

Treatment of RGC-5 cells with the broadspectrum protein kinase inhibitor staurosporine also differentiates them, but in a different way (Frassetto et al., 2006). The mechanism by which staurosporine induces RGC-5 cell differentiation is

281

different from staurosporine differentiation of other cell types. It is unlikely to be a result of apoptosis because staurosporine, a known apoptosis inducer, does not activate apoptotic cascade in RGC-5 cells. This is an important distinction because differentiation resulting in apoptosis would not be useful for studying RGC pathophysiology. Staurosporine induces RGC-5 cells to differentiate, express neurites, and become postmitotic. Staurosporine also induces electrophysiological changes that are in the same direction as mature RGC because both cells have large voltagegated conductance.

Staurosporine-differentiated RGC-5 cells differ in significant ways from primary cultured RGCs. Staurosporine differentiation is transcription independent and results in cells that are viable in the absence of any neurotrophic factor support, unlike normally differentiated RGCs. Neurotrophic factor dependence would be a necessary component for reproducing functional connectivity of neurons to the central nervous system, which is the goal of in vivo application of neuronal stem cells.

A third method of differentiation is with histone deacetylase (HDAC) inhibition. We have studied the relation between histone acetylation and the differentiation and survival of RGC-5 cells and compared it with the transcription-independent differentiation induced by staurosporine (Schwechter et al., 2007). Trichostatin (TSA) is a potent, specific, and well-characterized class 1 and class 2 HDAC inhibitor. TSA causes significant differentiation and neuritogenesis in RGC-5 cells. Differences between HDAC inhibition and staurosporine differentiation include the proportion of differentiated cells, cell viability, cell morphology, and transcriptional dependence. Also, treatment of RGC-5 cells with TSA resulted in RGC-5 cells that are neurotrophic factor dependent, unlike cells treated with staurosporine. Interestingly, HDAC inhibition also increases the sensitivity of RGC-5 cells to differentiation by very low concentrations of staurosporine (Schwechter et al., 2007).

Although not strictly a fourth method of differentiation, Harvey and Chintala studied the effect of plasminogen activators and their inhibition on staurosporine-treated RGC-5 cells (Harvey and Chintala, 2007). Plasmin aids in the

282

elongation process of newly formed neurites by degrading the extracellular matrix. Plasminogen activators convert plasminogen to plasmin. However, under certain circumstances, plasminogen activators promote cell death. Undifferentiated RGC-5 cells do not express the plasminogen activators, tPA and uPA. When treated with staurosporine, RGC-5 expression of those two plasminogen activators is observed. At a high staurosporine concentration, there is an increase in tPA and uPA, but there is also an increase in cell death and shorter neurites. When RGC-5 cells are treated with staurosporine and plasminogen activator inhibitors, there is a decrease in plasminogen activator proteolytic activity and cell death, and an increase in neurite outgrowth. Thus, differentiation of RGC-5 cells with staurosporine induces the expression of tPA and uPA, and these plasminogen activators cause RGC-5 death. By inhibiting this process, the neuritic tree can be stabilized and survival enhanced (Harvey and Chintala, 2007).

RGC-5 cell neurites

Much work has been done to characterize the factors guiding axonal pathfinding in RGCs to appropriate sites in the brain (Oster et al., 2004). Studies have also characterized factors controlling the formation of the dendritic arbor and its stratification in the retina during development, including cell density (Troilo et al., 1996) and neurotrophin levels (Lom et al., 2002). However, none of these studies monitored neurite development over time in a single RGC. Purified RGCs have been cultured and their neurite outgrowth has been studied intensively, but these cells have undergone injury to their existing neuritis in the process of purification.

RGC-5 cells express what appears to be axon and dendrites (Lieven et al., 2007). Microtubuleassociated proteins are a family of proteins responsible for microtubule stabilization and organization. Microtubule-associated protein 2 (MAP2) is particularly involved in cytoskeletal changes associated with neuronal differentiation (Caceres et al., 1986). MAP2 exists in several isoforms, the most prevalent being MAP2a,

MAP2b, and MAP2c. MAP2a and MAP2b are present in dendrites of mature neurons (Bernhardt and Matus, 1984). MAP2c is present in developing neurons (Meichsner et al., 1993) but not in mature neurons. Tau is another microtubule-associated protein expressed in differentiated neurons, expressed exclusively in the axon in vivo (Binder et al., 1985) and in the soma of cultured cells. The expression of tau characteristically presents as a gradient, with greater amounts at the distal axon (Kempf et al., 1996). Growth-associated protein 43 (GAP43) is neuronal protein expressed in neurite growth cones, specifically those of axons (Goslin et al., 1988).

Our studies confirmed MAP2c expression in some neurites of staurosporine-treated RGC-5 cells. The expression of GAP43 in growth cones and the presence of a tau gradient confirm the development of axons and the establishment of neuronal polarity in these cells. However, the use of staurosporine-differentiated RGC-5 cells as a model for neurite formation in RGCs has potential shortcomings. It is unclear whether this staurosporine differentiation program is similar to what occurs during differentiation of primary RGCs. Also, the number of axons is low (Lieven et al., 2007).

Advantages and disadvantages of culture models

The processes studied with cell culture models appear to be so distinct from those associated with optic neuropathies as to make them less helpful for studying disease. Yet there are many features of cell culture that cannot be replicated in animal models. In cell culture models, there is an ability to control a cell’s exposure to specific chemical factors, drugs, interactions with different cell types, and changes in the extracellular milieu. Multiple conditions can be studied in parallel within the same experiment when it would be near impossible to do so in situ within an animal’s eye. Results of those studies may sometimes be extrapolated to in vivo conditions; identification of brain-derived neurotrophic factor (BDNF) as a survival factor for cultured RGCs is a good example.